We have assessed the effect of combining extradural bupivacaine and fentanyl in 60 orthopaedic patients who received 0.125% bupivacaine (bupivacaine group), fentanyl 5 micrograms ml-1 (fentanyl group), or 0.125% bupivacaine combined with fentanyl 5 micrograms ml-1 (combined group), delivered by patient-controlled extradural analgesia for 24 h via a lumbar extradural catheter. Adding bupivacaine to fentanyl reduced mean (SD) fentanyl administration from 117 (46) ml to 89 (42) ml (P < 0.005). Adding fentanyl to bupivacaine reduced mean bupivacaine administration from 113 (46) ml to 89 (42) ml (P < 0.05). There was no significant difference between the groups in pain, nausea, motor block, pruritus or sedation. No patient had a ventilatory frequency less than 10 b.p.m. Hypotension (systolic AP < 100 mm Hg) occurred in two of 20 patients in the fentanyl group, compared with eight of 19 in the bupivacaine group and 10 of 21 in the combined group. The mean total volume of extradural solution administered was greater after knee replacement (126 (46) ml) than after hip replacement (84 (35) ml) (P < 0.001). The mean pain score was greater also for knee replacement (16 (10) mm) than for hip replacement (10 (9) mm) (P < 0.05). We conclude that extradural bupivacaine and fentanyl were additive in their analgesic actions, resulting in decreased requirements of each individual agent. Knee replacement was found also to be more painful than hip replacement after operation.